2016-03-22 00:23:37 +03:00
|
|
|
/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
|
2005-05-24 02:06:50 +04:00
|
|
|
/*
|
2005-11-05 22:57:48 +03:00
|
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
2007-07-12 02:21:04 +04:00
|
|
|
* Copyright (c) 2004-2007 The University of Tennessee and The University
|
2005-11-05 22:57:48 +03:00
|
|
|
* of Tennessee Research Foundation. All rights
|
|
|
|
* reserved.
|
2015-06-24 06:59:57 +03:00
|
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
2005-05-24 02:06:50 +04:00
|
|
|
* University of Stuttgart. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
2007-10-15 17:10:20 +04:00
|
|
|
* Copyright (c) 2006-2007 Sun Microsystems, Inc. All rights reserved.
|
2013-05-01 16:26:16 +04:00
|
|
|
* Copyright (c) 2008-2013 Cisco Systems, Inc. All rights reserved.
|
2017-01-06 07:12:16 +03:00
|
|
|
* Copyright (c) 2015-2017 Research Organization for Information Science
|
2015-04-15 09:14:57 +03:00
|
|
|
* and Technology (RIST). All rights reserved.
|
2018-02-07 19:49:36 +03:00
|
|
|
* Copyright (c) 2016-2018 Los Alamos National Security, LLC. All rights
|
2016-03-22 00:23:37 +03:00
|
|
|
* reserved.
|
2005-05-24 02:06:50 +04:00
|
|
|
* $COPYRIGHT$
|
2015-06-24 06:59:57 +03:00
|
|
|
*
|
2005-05-24 02:06:50 +04:00
|
|
|
* Additional copyrights may follow
|
2015-06-24 06:59:57 +03:00
|
|
|
*
|
2005-05-24 02:06:50 +04:00
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
#include "opal_config.h"
|
2005-05-24 02:06:50 +04:00
|
|
|
#include <stdio.h>
|
|
|
|
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "opal/mca/mca.h"
|
2009-02-14 05:26:12 +03:00
|
|
|
#include "opal/util/output.h"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "opal/mca/base/base.h"
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
#include "opal/mca/btl/btl.h"
|
|
|
|
#include "opal/mca/btl/base/base.h"
|
2005-05-24 02:06:50 +04:00
|
|
|
|
2016-03-22 00:23:37 +03:00
|
|
|
mca_base_var_enum_flag_t *mca_btl_base_flag_enum = NULL;
|
|
|
|
mca_base_var_enum_flag_t *mca_btl_base_atomic_enum = NULL;
|
|
|
|
|
|
|
|
mca_base_var_enum_value_flag_t mca_btl_base_flag_enum_flags[] = {
|
|
|
|
{MCA_BTL_FLAGS_SEND, "send", 0},
|
|
|
|
{MCA_BTL_FLAGS_PUT, "put", 0},
|
|
|
|
{MCA_BTL_FLAGS_GET, "get", 0},
|
|
|
|
{MCA_BTL_FLAGS_SEND_INPLACE, "inplace", 0},
|
|
|
|
{MCA_BTL_FLAGS_SIGNALED, "signaled", 0},
|
|
|
|
{MCA_BTL_FLAGS_ATOMIC_OPS, "atomics", 0},
|
|
|
|
{MCA_BTL_FLAGS_ATOMIC_FOPS, "fetching-atomics", 0},
|
|
|
|
{MCA_BTL_FLAGS_SINGLE_ADD_PROCS, "static", 0},
|
|
|
|
{MCA_BTL_FLAGS_CUDA_PUT, "cuda-put", 0},
|
|
|
|
{MCA_BTL_FLAGS_CUDA_GET, "cuda-get", 0},
|
|
|
|
{MCA_BTL_FLAGS_CUDA_COPY_ASYNC_SEND, "cuda-async-send", 0},
|
|
|
|
{MCA_BTL_FLAGS_CUDA_COPY_ASYNC_RECV, "cuda-async-recv", 0},
|
|
|
|
{MCA_BTL_FLAGS_FAILOVER_SUPPORT, "failover", 0},
|
|
|
|
{MCA_BTL_FLAGS_NEED_ACK, "need-ack", 0},
|
|
|
|
{MCA_BTL_FLAGS_NEED_CSUM, "need-csum", 0},
|
|
|
|
{MCA_BTL_FLAGS_HETEROGENEOUS_RDMA, "hetero-rdma", 0},
|
2018-02-07 19:49:36 +03:00
|
|
|
{MCA_BTL_FLAGS_RDMA_FLUSH, "rdma-flush", 0},
|
2016-03-22 00:23:37 +03:00
|
|
|
{0, NULL, 0}
|
|
|
|
};
|
|
|
|
|
|
|
|
mca_base_var_enum_value_flag_t mca_btl_base_atomic_enum_flags[] = {
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_ADD, "add", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_AND, "and", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_OR, "or", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_XOR, "xor", 0},
|
2016-05-25 19:39:34 +03:00
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_LAND, "land", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_LOR, "lor", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_LXOR, "lxor", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_SWAP, "swap", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_MIN, "min", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_MAX, "max", 0},
|
2016-03-22 00:23:37 +03:00
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_CSWAP, "compare-and-swap", 0},
|
|
|
|
{MCA_BTL_ATOMIC_SUPPORTS_GLOB, "global"},
|
|
|
|
{0, NULL, 0}
|
|
|
|
};
|
|
|
|
|
2015-04-15 09:14:57 +03:00
|
|
|
mca_btl_active_message_callback_t mca_btl_base_active_message_trigger[MCA_BTL_TAG_MAX] = {{0}};
|
2008-01-15 08:32:53 +03:00
|
|
|
|
2005-05-24 02:06:50 +04:00
|
|
|
/*
|
2005-06-30 06:08:24 +04:00
|
|
|
* mca_btl_base_descriptor_t
|
2005-05-24 02:06:50 +04:00
|
|
|
*/
|
|
|
|
|
2005-06-30 06:08:24 +04:00
|
|
|
static void mca_btl_base_descriptor_constructor(mca_btl_base_descriptor_t* des)
|
2005-05-24 02:06:50 +04:00
|
|
|
{
|
2015-01-06 00:55:15 +03:00
|
|
|
des->des_segments = NULL;
|
|
|
|
des->des_segment_count = 0;
|
2005-05-24 02:06:50 +04:00
|
|
|
des->des_cbfunc = NULL;
|
|
|
|
des->des_cbdata = NULL;
|
|
|
|
des->des_flags = 0;
|
|
|
|
}
|
|
|
|
|
2005-06-30 06:08:24 +04:00
|
|
|
static void mca_btl_base_descriptor_destructor(mca_btl_base_descriptor_t* des)
|
2005-05-24 02:06:50 +04:00
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
OBJ_CLASS_INSTANCE(
|
2005-06-30 06:08:24 +04:00
|
|
|
mca_btl_base_descriptor_t,
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_item_t,
|
2005-06-30 06:08:24 +04:00
|
|
|
mca_btl_base_descriptor_constructor,
|
|
|
|
mca_btl_base_descriptor_destructor);
|
2005-05-24 02:06:50 +04:00
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The following file was created by configure. It contains extern
|
|
|
|
* statements and the definition of an array of pointers to each
|
|
|
|
* component's public mca_base_component_t struct.
|
|
|
|
*/
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
#include "opal/mca/btl/base/static-components.h"
|
2005-08-09 21:49:39 +04:00
|
|
|
#include "btl_base_error.h"
|
2005-05-24 02:06:50 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Global variables
|
|
|
|
*/
|
2005-06-30 06:08:24 +04:00
|
|
|
char* mca_btl_base_include = NULL;
|
|
|
|
char* mca_btl_base_exclude = NULL;
|
2006-09-27 17:24:42 +04:00
|
|
|
int mca_btl_base_warn_component_unused = 1;
|
2015-04-15 09:14:57 +03:00
|
|
|
opal_list_t mca_btl_base_modules_initialized = {{0}};
|
2008-10-28 21:29:57 +03:00
|
|
|
bool mca_btl_base_thread_multiple_override = false;
|
2005-05-24 02:06:50 +04:00
|
|
|
|
2013-03-28 01:17:31 +04:00
|
|
|
static int mca_btl_base_register(mca_base_register_flag_t flags)
|
2013-03-28 01:09:41 +04:00
|
|
|
{
|
|
|
|
/* Override the per-BTL "don't run if THREAD_MULTIPLE selected"
|
|
|
|
embargo? */
|
|
|
|
mca_btl_base_thread_multiple_override = false;
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
(void) mca_base_var_register("opal", "btl", "base", "thread_multiple_override",
|
2013-03-28 01:09:41 +04:00
|
|
|
"Enable BTLs that are not normally enabled when MPI_THREAD_MULTIPLE is enabled (THIS IS FOR DEVELOPERS ONLY! SHOULD NOT BE USED BY END USERS!)",
|
|
|
|
MCA_BASE_VAR_TYPE_BOOL, NULL, 0,
|
|
|
|
MCA_BASE_VAR_FLAG_INTERNAL,
|
|
|
|
OPAL_INFO_LVL_9,
|
|
|
|
MCA_BASE_VAR_SCOPE_READONLY,
|
|
|
|
&mca_btl_base_thread_multiple_override);
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
(void) mca_base_var_register("opal", "btl", "base", "include", NULL,
|
2013-03-28 01:09:41 +04:00
|
|
|
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0,
|
|
|
|
OPAL_INFO_LVL_9,
|
|
|
|
MCA_BASE_VAR_SCOPE_READONLY,
|
|
|
|
&mca_btl_base_include);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
(void) mca_base_var_register("opal", "btl", "base", "exclude", NULL,
|
2013-03-28 01:09:41 +04:00
|
|
|
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0,
|
|
|
|
OPAL_INFO_LVL_9,
|
|
|
|
MCA_BASE_VAR_SCOPE_READONLY,
|
|
|
|
&mca_btl_base_exclude);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
(void) mca_base_var_register("opal", "btl", "base", "warn_component_unused",
|
2013-03-28 01:09:41 +04:00
|
|
|
"This parameter is used to turn on warning messages when certain NICs are not used",
|
|
|
|
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
|
|
|
|
OPAL_INFO_LVL_9,
|
|
|
|
MCA_BASE_VAR_SCOPE_READONLY,
|
|
|
|
&mca_btl_base_warn_component_unused);
|
|
|
|
|
2016-03-22 00:23:37 +03:00
|
|
|
(void) mca_base_var_enum_create_flag ("btl_flags", mca_btl_base_flag_enum_flags, &mca_btl_base_flag_enum);
|
2017-01-06 07:12:16 +03:00
|
|
|
(void) mca_base_var_enum_register("opal", "btl", "base", "btl_flags",
|
|
|
|
&mca_btl_base_flag_enum);
|
2016-03-22 00:23:37 +03:00
|
|
|
(void) mca_base_var_enum_create_flag ("btl_atomic_flags", mca_btl_base_atomic_enum_flags, &mca_btl_base_atomic_enum);
|
2017-01-06 07:12:16 +03:00
|
|
|
(void) mca_base_var_enum_register("opal", "btl", "base", "btl_atomic_flags",
|
|
|
|
&mca_btl_base_atomic_enum);
|
2016-03-22 00:23:37 +03:00
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
return OPAL_SUCCESS;
|
2013-03-28 01:09:41 +04:00
|
|
|
}
|
|
|
|
|
2005-05-24 02:06:50 +04:00
|
|
|
/**
|
|
|
|
* Function for finding and opening either all MCA components, or the one
|
|
|
|
* that was specifically requested via a MCA parameter.
|
|
|
|
*/
|
2013-03-28 01:17:31 +04:00
|
|
|
static int mca_btl_base_open(mca_base_open_flag_t flags)
|
2005-05-24 02:06:50 +04:00
|
|
|
{
|
2013-03-28 01:17:31 +04:00
|
|
|
int ret;
|
2007-01-11 00:24:35 +03:00
|
|
|
|
2013-03-28 01:09:41 +04:00
|
|
|
/* Open up all available components */
|
2015-06-24 06:59:57 +03:00
|
|
|
|
|
|
|
if (OPAL_SUCCESS !=
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
(ret = mca_base_framework_components_open(&opal_btl_base_framework, flags))) {
|
2013-03-28 01:17:31 +04:00
|
|
|
return ret;
|
2013-03-28 01:09:41 +04:00
|
|
|
}
|
2005-05-24 02:06:50 +04:00
|
|
|
|
2005-06-30 06:08:24 +04:00
|
|
|
/* Initialize the list so that in mca_btl_base_close(), we can
|
2005-05-24 02:06:50 +04:00
|
|
|
iterate over it (even if it's empty, as in the case of
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
opal_info) */
|
2005-05-24 02:06:50 +04:00
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
OBJ_CONSTRUCT(&mca_btl_base_modules_initialized, opal_list_t);
|
2005-05-24 02:06:50 +04:00
|
|
|
|
2013-04-30 05:15:52 +04:00
|
|
|
/* get the verbosity so that BTL_VERBOSE will work */
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
mca_btl_base_verbose = opal_output_get_verbosity(opal_btl_base_framework.framework_output);
|
2013-04-30 05:15:52 +04:00
|
|
|
|
2005-05-24 02:06:50 +04:00
|
|
|
/* All done */
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
return OPAL_SUCCESS;
|
2005-05-24 02:06:50 +04:00
|
|
|
}
|
2013-03-28 01:17:31 +04:00
|
|
|
|
|
|
|
static int mca_btl_base_close(void)
|
|
|
|
{
|
2013-05-01 16:26:16 +04:00
|
|
|
mca_btl_base_selected_module_t *sm, *next;
|
2013-03-28 01:17:31 +04:00
|
|
|
|
|
|
|
#if 0
|
|
|
|
/* disable event processing while cleaning up btls */
|
|
|
|
opal_event_disable();
|
|
|
|
#endif
|
|
|
|
/* Finalize all the btl components and free their list items */
|
|
|
|
|
2013-05-01 16:26:16 +04:00
|
|
|
OPAL_LIST_FOREACH_SAFE(sm, next, &mca_btl_base_modules_initialized, mca_btl_base_selected_module_t) {
|
2013-03-28 01:17:31 +04:00
|
|
|
/* Blatently ignore the return code (what would we do to recover,
|
|
|
|
anyway? This component is going away, so errors don't matter
|
|
|
|
anymore) */
|
|
|
|
|
|
|
|
sm->btl_module->btl_finalize(sm->btl_module);
|
2013-05-01 16:26:16 +04:00
|
|
|
opal_list_remove_item(&mca_btl_base_modules_initialized, &sm->super);
|
2013-03-28 01:17:31 +04:00
|
|
|
free(sm);
|
|
|
|
}
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
(void) mca_base_framework_components_close(&opal_btl_base_framework, NULL);
|
2013-03-28 01:17:31 +04:00
|
|
|
|
2013-05-01 16:26:16 +04:00
|
|
|
OBJ_DESTRUCT(&mca_btl_base_modules_initialized);
|
|
|
|
|
2013-03-28 01:17:31 +04:00
|
|
|
#if 0
|
|
|
|
/* restore event processing */
|
|
|
|
opal_event_enable();
|
|
|
|
#endif
|
|
|
|
/* All done */
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
return OPAL_SUCCESS;
|
2013-03-28 01:17:31 +04:00
|
|
|
}
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
MCA_BASE_FRAMEWORK_DECLARE(opal, btl, "Byte Transport Layer", mca_btl_base_register,
|
2013-03-28 01:17:31 +04:00
|
|
|
mca_btl_base_open, mca_btl_base_close, mca_btl_base_static_components,
|
|
|
|
0);
|